Making Sense of (Ultra) Low Cost Flights

Making Sense of (Ultra) Low Cost Flights
Vertical Differentiation in Two-Sided Markets
Luigi Serio∗
Piero Tedeschi∗
Giovanni Ursino∗
February 16, 2015
very preliminary draft
Abstract
The business model of low cost carriers is now well established and accounts for a large share of
western civil aviation, particularly in Europe. To understand why it has proven so successful, we
develop a theoretical model which exploits the two-sided nature of flights as connectors of supply and
demand for goods and services other than traveling itself across physical space. Carriers offer flights of
different quality and may sign agreements with suppliers of goods and services at destination so as to
subsidize and foster demand from the carriers’ travelers as in standard two-sided markets. Customertravelers care about home and destination consumption and about the flight’s quality. Hence, beyond
the thickness of the connected sides of the market, the quality of the airline-platform has an intrinsic
value to travelers. We show that only cash constrained travelers fly with low cost airlines, while
no-frills carriers are more likely to act as a platform than legacy airlines. We study the impact on
the equilibrium market structure of the airline industry of several features, such as how competitive
is the destination market, or the extent to which home and destination consumption are substitutes.
JEL Classification Numbers: L1, L2
Keywords: vertical differentiation, two-sided markets, air travel, low cost flights
∗
Catholic University of Milan. Contact: [email protected].
1
Introduction
Almost all Europeans and most Americans are familiar with low cost flying. While low cost airlines
exist since the ’70s,1 the phenomenon has constantly grown. At least since the late ’90s it has reached
a mass market status into what is now considered an established but still growing industry. This is
particularly true of the European commercial aviation: with the emergence of budget airlines in the
late ’90s the growth trend of established network carriers has stopped. Today, full service carriers still
handle approximately the same demand for air travel as in 2000, while their relative market share has
decreased. In contrast, low cost carriers have grown at high double digit rates and captured large parts
of the market. They have expanded their market share from 5% in 2001 to 32% in 2008. In some
European countries, low cost carriers even dominate the market. In Spain, they account for 50% of the
total international capacity offered, in Poland even for 52%. Also in other economies, budget carriers
have been able to expand their market share in an impressive way over the last decade, which puts
severe pressure on established network carriers. In Germany, Europe’s largest economy, low cost airlines
operate 29% of the international and 44% of the domestic flights.2
While low cost companies were growing steadily, the 2000s witnessed a sharp increase in the demand
for air travel in Europe, which jumped by around 47% between 2000 and 2008.3 To some extent this may
be due to a higher flying frequency for some travelers but that does not explain the entire picture. The
evidence suggests that the entrance of low cost carriers into the industry has made access to air traveling
possible to lower income people for whom traveling with established carriers was not affordable.4 A
considerable share of the new traffic is made up of tourists traveling inside Europe.5 At least until
recently, attracting this ‘new demand’ has been the core business of budget airlines.
The recognized champion of the low cost flying saga is Ryanair, which alone has a market share of
short haul passengers in Europe around 14%.6 Such an impressive score is accompanied by announcements like that released by Ryanair chief executive Michael O’Leary in November 5, 2007: “It’s our
ultimate ambition to get to a stage where the fare is free.”7 That interview focused on cost containment
1
During the ’60s and the ’70s Loftleidir pioneered as a low fare service airline across the North-Atlantic flying into
Luxembourg “the heart of Europe.” The airline became very popular among college students traveling abroad and soon
became know as “The Hippie Airline” flying among others former US president Bill Clinton. The first fully low cost airline
is generally considered to be the American company Southwest which launched in 1971, with the then revolutionary concept
that you could lower the cost of ticket prices by eliminating some of the extras and therefore save passengers money.
2
Future Scenarios for the European Airline Industry, a 2010 Report of the Center for Scenario Planning, Roland Berger
Research Unit and HHL - Leipzig Graduate School of Management.
3
See Footnote 2.
4
“In the 1950s flying was a privilege enjoyed by only the wealthiest. The costs of flying were simply too high for most
ordinary folk. In 1952 a London-to-Scotland return flight would set the average Englishman back a week’s wages; a trip to
New York might require saving up for five months. But in 2013 flying is a mass market, due in no small part to the growth of
“no-frills” airlines offering flights at very low prices.” from The Economist website at www.economist.com/blogs/economistexplains/2013/10/economist-explains-13.
5
See, e.g, the UK Civil Aviation Authority report “Demand for Outbound Leisure Air Travel and its Key Drivers”
(December 2005) available at www.caa.co.uk/docs/5/erg elasticity study.pdf.
6
Ryanair share of seats among all carriers in period April 29 – May 5, 2013. See www.centreforaviation.com/analysis/
ryanair-europes-lowest-cost-producer-wins-again-reporting-record-profit-of-eur569-million-110543.
7
See www.dailymail.co.uk/news/article-491907/No-cost-flights-Ryanair–passengers-incur-costs.html.
2
as Ryanair was about to introduce a tightening of the pay-for-frills policy with the doubling of check-in
as well as baggage fees. Indeed, cost minimization is the backbone strategy of every low cost airline:
saving on ‘frills’ to lower prices and attract low budget travelers. However, Ryanair’s average fare in
2013 (¤48) was by far the lowest, with the second cheapest (Easyjet, ¤82) charging 71% higher.8 Which
allows Ryanair to define itself as a (ultra) low cost carrier. Can this performance be explained just as a
result of cost reduction? Clearly not.
A distinguished feature of its business strategy is the way Ryanair deals with its more than seventy
‘bases’ around Europe.9 Among the airports in which Ryanair operates, the bases are those hosting the
carrier’s fleet. More importantly, most bases are almost exclusively operated by Ryanair, which is by far
the first carrier of the airport. This is not by chance: Ryanair carefully chooses its bases targeting minor
airports situated not far from attractive locations. The small size of such airports — at least before the
arrival of Ryanair — grants the carrier a bold stance when bargaining the terms of its operations. The
convenient location grants Ryanair a sustained demand, perhaps because next to an important touristic
attraction. The one described so far is just an aggressive cost minimizing/demand enhancing strategy.
Ryanair, however, is the only carrier which goes (way) beyond that: realizing that many of its
passengers are leisure travelers and likely customers of destination goods and services (hotels and touristic
services, local food products, fashion garment, etc.) the carrier actively exploits its role of connector
between demand and supply to extract part of (and sometimes most of) the potential gains from trade.
In fact, to some extent all air carriers build networks which create exchange opportunities — i.e. network
externalities. But only Ryanair actively exploits this platform cöté of air traveling. To see how, one
has to look at the contracts agreed to by Rayanair when the airline opens a base. Such contracts
are typically signed with the companies managing the airport but most often involve local authorities,
business representatives such as chambers of commerce and, more generally, ‘destination stakeholders’.
Some examples of destination stakeholders are: the Oriocenter Shopping Center10 in Milan Orio al Serio,
Airgest and Regione Sicilia in Trapani Birgi, Regione Puglia in Brindisi, Cataluña and Costa Brava hotels
in Reus and Gerona,11 etc.
The agreements between the (ultra) low cost carrier and destination stakeholders usually stipulate
that the latter pay Ryanair an amount varying with the number of passengers that the carrier commits
to fly at destination. For instance, Trapani paid ¤20 millions in five years and passengers soared from
533 thousands in 2008 to 1.2 millions in 2012.12 In some cases the contractual relationship is mutually
advantageous and sustainable, but it frequently happens that the terms disproportionately favor Ryanair:
while, for instance, Orio al Serio and Brindisi are success stories, the airport of Verona — which paid
8
Source: latest published company year end information, as reported in the Full Year Results 2013, Ryanair.
The opening of the 71st base (Bratislava) was announced on Nov 13, 2014 on the Ryanair Website. In December 2013
Ryanair had 57 bases according to the Full Year Results 2013.
10
More information at www.oriocenter.it.
11
See (in Spanish) www.elperiodicomediterraneo.com/noticias/castellon/cataluna-paga-46-millones-ryanair-traer-turistas
-reus-girona 728694.html.
12
For concise information on Italian airports dealing with Ryanair see (in Italian) https://it.finance.yahoo.com/foto/gliaiutini-di-stato-a-ryanair-slideshow.
9
3
the carrier ¤24/passenger until recently — got close to bankruptcy.13
This paper is about the economics underlying the business model of Ryanair. It aims at explaining
the success of the (ultra) low cost carrier, its impact on the industry and the consequences for the
consumption habits of millions of low–to–mid income people. It does so by recognizing and modeling
the unique feature of Ryanair’s strategy: not just competing on quality as a normal low cost carrier, but
actively exploiting the network externalities inherent in moving people across markets. Ryanair acts as a
platform connecting demand and supply located in different countries and extracts part of the generated
surplus to keep its fares at otherwise unprofitably low levels. It therefore competes on quality with
other carriers but, at the same time, generates profits by selling its platform services in many two-sided
markets around Europe, each associated to a specific network externality characteristic of each route.
We deem natural to model the airline industry as a vertically differentiated duopoly in which two
airlines, a low quality (low cost) carrier and a high quality (full service) one, compete to attract potential
travelers on a given route as in standard models. Customer-travelers have different income levels and care
primarily about the flight’s quality and, secondly, about goods’ consumption. They can purchase goods
both at home and, if they travel, at destination, where they can source from a number of shops. These
destination businesses are collectively represented by a local stakeholder (chamber of commerce, regional
council, shopping mall, etc.). We depart from the standard vertical differentiation setup assuming that
carriers have the opportunity to propose a contract for operating the route to the local stakeholder.
The latter can strike a deal with just one carrier — that is, contracts are exclusive. The destination
stakeholder, however, sells goods and services to the travelers brought at destination by both carriers.
Hence, while traveling with either carrier generates positive network externalities on travelers — i.e.,
being able to purchase at destination — only one carrier can price on both sides of the market as a
proper platform, whereas the other prices just on the traveler’s side. Put differently, exclusivity implies
that only one carrier can internalize the network externalities it generates. We study and compare three
scenarios: a standard competition benchmark in which no carrier deals with the local stakeholder, one
in which the stakeholder deals with the full service carrier and one in which he deals with the low cost
carrier.
We find that the optimal contract between the local stakeholder and a carrier is a two-part tariff
prescribing a per-passenger payment to the carrier in exchange of a fixed amount to the stakeholder.
Not surprisingly, the stakeholder finds always optimal to deal with a carrier rather than not signing any
contract — i.e., the competition benchmark is never an equilibrium when contracting is possible. More
importantly, we show that the overall demand for air traveling depends on the level of low cost carrier
tariff and increases when the latter decreases. Indeed, we find that demand is larger when a carrier
deals with the stakeholder and, in particular, it is largest when the carrier is a low cost one, confirming
the evidence from low cost airlines’ emergence and passengers’ data. Moreover, travelers flying low cost
are cash-constrained — i.e., they would prefer to travel on a full service flight — while very low-income
consumers prefer not to fly at all — i.e., the marginal passenger is not cash-constrained.
13
See Footnote 12.
4
Next, we find conditions under which contracting with the low cost carrier is more efficient than
contracting with the full service carrier. It turns out that dealing with the low cost airline is more
efficient whenever the cost of consumption is high relative to the cost of flying: when this is so, because
the tariff reduction is stronger under low cost–stakeholder contracting, dealing with the low cost generates
a larger increase in demand, as seen above, and allows consumers/travelers to spare more money to spend
at destination. We then verify how the parameters of the model such as the love for variety, the number
of goods’ varieties and their production costs affect the chances that the equilibrium contract is offered
to either carrier.
Our results closely track the strategy and market outcomes observed in practice. From a theoretical
point of view they are made possible by the combination of vertical differentiation and two-sidedness
into a single framework. The literature is vast on both aspects and reviewing it goes beyond the scope
of this work. See, however, Rochet and Tirole (2006) on two-sided markets and Tirole (1988) on vertical
differentiation. Our work lays at the intersection of these strands of the literature. To the best of our
knowledge, no other work shares a structure similar to ours. However, following Armstrong (2006),
many authors have studied competition between networks and the idea that competing platforms may
differentiate is not novel. Although we do not model network competition, it is worth mentioning those
works which have studied differentiation in such models. Argenziano (2008) presents a model in which
two ex-ante identical networks compete to attract agents who have imperfect information about the
network quality (a common value) as well as heterogeneous (private) valuations about the goods on
sale. She finds conditions for unique equilibria to emerge and shows that networks are suboptimally
differentiated because consumers fail to internalize externalities due to the asymmetric information
structure of the model. Our model differs in several ways: carriers’ qualities — i.e., the common value —
are common knowledge; consumers are heterogeneous with respect to their income levels, not preferences;
both carriers compete on the air traveling market but only one uses platform pricing. Gabszewicz and
Wauthy (2014) model competition between two vertically differentiated platforms — while Ribeiro,
Correia-da-Silva, and Resende (2014) build on their framework. In these works a platform’s quality is
endogenous and is higher the larger the market share and the associated network externalities. Their
platforms compete to attract users as in classical models and have no value per se. To the contrary,
because in our framework consumers care directly about it, airlines compete on the intrinsic quality of
their flights, independently of network externalities. These are internalized on top of and interact with
the standard competitive environment of a vertically differentiated duopoly.
The paper is organized as follows: Section 2 presents the model. Section 3 analyses optimal consumer’s behavior, studies monopolistic competition between destination shops and derives flights’ demand as a function of the carrier’s pricing decisions. Section 4 studies the standard competition scenario,
the scenario in which the low cost carrier deals with the local stakeholder and than in which the full
service does, and compare the equilibrium outcomes. Section 5 characterizes the market structure which
emerges endogenously in equilibrium and its determinants. Section 6 concludes.
5
2
Model
Players. Two airlines must decide how to operate a route to a given location providing services of
different qualities: airline F is a Full Service carrier committed to high-quality standards, airline L is
a Low Cost carrier with a no-frills policy. Carrier F offers a quality QF and charges a tariff TF per
passenger while carrier L offers quality QL < QF and charges TL . For tractability reasons we take
qualities as exogenous, QL is normalized to 0 without loss of generality, and we assume that carrier F
sustains no cost for meeting quality standard QF .14
Located near the destination airport are M ≥ 3 businesses — they could be hotels, museums, etc.
but for the remaining of the paper we will call them shops and, to name them collectively, we will adopt
the Mall example, using alternatively the terms shops and Mall. Each traveler k may buy a quantity
qik ≥ 0 of the good provided by shop i at price pi . Shops are naturally interested in airlines’ decisions
insofar as passengers are potential buyers. In order to attract buyers, the shops — which, for the sake
of simplicity, create the ‘Mall’ to deal with airline companies15 — may be willing to subsidize airline I
with a transfer SI per traveler carried to destination. To partially compensate for this, optimal contracts
between airlines and the Mall may have a two-part tariff structure, prescribing a fixed transfer ZI to the
Mall, which is evenly split between shops. In this case each shop subsidizes airline I with a payment
sI = SI /M per traveler carried by airline I and receives a fixed payment zI = ZI /M .16 Goods have
constant unit cost c > 0 to each shop. Beyond goods at destination, whether he flies or not, a potential
traveler k may consume a quantity q0k ≥ 0 of a numéraire good at home at the normalized price p0 = 1.
Finally, there is a unit-mass of travelers indexed by k and characterized by a uniformly distributed
income Ik ∼ U [0, 1]. Travelers use all their income for flying and buying goods, i.e. they derive no utility
from money per se and care only about the flight’s quality and shopping at home and at destination.
Building on Dixit and Stiglitz (1977), the utility function of traveler k when flying with carrier I is a
nested CES on consumption goods with the addition of a quasi-linear component on flight’s quality

ϕ
+
UIk (QIk , q0k , q1k , ..., qM k ) = QIk + q0k
M
X
! ϕρ  ϕ1
ρ
qik

I = F, L
(1)
i=1
where the CES parameters ϕ ∈ (0, 1) and ρ ∈ (0, 1) reflect consumers’ taste for variety: between home
14
While the model is robust to the introduction of a cost of quality — as long as it keeps the Full Service carrier in
business —, the zero costs assumption simplifies the analysis without sacrificing any intuition. The only drawback is that
the profits of the Full Service carrier are trivially higher than those of the Low Cost carrier in equilibrium. A complete
analysis is available upon request.
15
We assume such an association exists and we justify this assumption on two grounds: from an empirical point of view
associations like these are common in practice (e.g. the Oriocenter Shopping Center, in Milan Orio al Serio or Chambers of
Commerce, Hotels’ in Gerona and Reus, etc. See Footnotes 11 and 12); from a theoretical perspective, instead, coordination
incentives are typically strong in contexts as the one described here, as it will be clear in the remainder of the paper.
16
In most of the analysis we will refer to contracts specifying the Mall-level transfers SI and ZI , but, particularly when
dealing with shops’ optimal behavior and occasionally elsewhere, we will use the shop-level notation sI and zI .
6
and destination goods, ϕ, and between goods at destination, ρ.17 The k subscript denotes the choice
of consumer k endowed with income Ik . The utility of a consumer who does not travel depends on the
numéraire good alone and is
k
UN
(q0k ) = q0k .
Timing. The structure of the airline industry game is as follows:
t = 1 Carriers set the quality levels QI ∈ {0, QF }.
t = 2 Carriers set fares TI and propose contracts (sI , zI ) to the Mall.
t = 3 The Mall decides whether and which contract to accept and shops set prices pi , i = 1, .., M .
t = 4 Consumers decide whether to fly and the amount of goods to purchase.
Equilibrium concept and strategies. The game is a sequential game with complete information and
the natural solution concept is Subgame Perfect Nash Equilibrium. We focus on pure strategies. The
actions available to carrier I are {QI , TI , SI , SI } with QI ∈ {0, QF }, TI , SI and ZI non-negative. The
action space of each shop is {‘Accept (sI , zI ) ’, ‘Not Accept (sI , zI ) ’, pi } for I = F, L and i = 1, ..., M
with pi ≥ 0. Finally, traveler k’s actions are {‘Not fly’, ‘Fly with F ’, ‘Fly with L’, q0k , ..., qM k }.
3
Consumers’ and shops’ behavior
Consumer’s behavior. Our equilibrium analysis proceeds by backward induction, starting from the
consumer’s decisions. There are three types of consumers: those who don’t fly, those who fly with F
and those who fly with L. Consumers who don’t fly spend all their income for the home good and get
k = I . The consumption decisions of those who fly are more nuanced. Let’s thus study the
utility UN
k
purchasing choices at destination of a traveler who has decided to fly with airline I ∈ {F, L}. Given
shops’ prices, the consumer maximizes (1) subject to the budget constraint
Ik − q0k − TI −
XM
i=1
pi qik ≥ 0.
(2)
The optimal demand of consumer k for the numéraire and destination good i is
Ik − TI
1 + P 1−τ
P σ−τ Ik − TI
=
pσi 1 + P 1−τ
q0k =
qik
17
(3)
i = 1, ..., M
(4)
1
The parameter ρ determines the elasticity of substitution between shops’ items, σ = 1−ρ
. A higher ρ implies a higher
1
substitutability in consumers’ preferences for destination goods, a lower ρ a higher taste for variety. Similarly, τ = 1−ϕ
is
the elasticity of substitution between the home numéraire q0 and the bundle of destination goods (q1 , ..., qM ).
7
where P ≡
P
M
1−σ
i=1 pi
1
1−σ
is the “destination price index” and σ =
1
1−ρ
> 1 and τ =
1
1−ϕ
> 1
are the elasticity of substitution between goods at destination and between home and destination goods
respectively.18 Clearly, the quantity of both home and destination goods purchased by traveler k increases
with his income net of traveling costs (‘net income’ hereafter).
Combining (1), (3) and (4), the utility of traveler k flying with I is
UIk = QI +
where Π ≡ 1 + P 1−τ
1
1−τ
Ik − TI
,
Π
(5)
is a “global price index” comprehending both home prices (11−τ ) and the
price index of destination goods (P 1−τ ) and does not exceed 1.19 Naturally, UIk increases in the flight’s
quality, QI , and in net real income,
Ik −TI
Π .
Having characterized goods’ consumption choices of consumer k, let’s now consider his flying decisions. Let’s begin with the choice to travel altogether. He will travel if, for at least one airline I, it
holds
k
UIk ≥ UN
⇐⇒
Ik ≥
TI − ΠQI
,
1−Π
(6)
while he will not travel if (6) is never satisfied. The flying condition above is stricter the higher the cost
of flying, TI , while it is met more easily when flying provides a higher utility, QI . More interestingly,
the effect of the price index Π on the flying decision depends on whether QI /TI — the flight’s marginal
utility per euro spent on flying — is greater or lower than 1: as Π increases, less (resp. more) consumers
will travel when QI < TI (resp. >). To understand why, notice that: i) the utility is linear in income
k = I ); and, ii) given the constant price of the numéreaire, p = 1, an
for those who don’t fly (UN
0
k
increase of Π is, in fact, an increase in P , the price of destination goods. Thus, when QI < TI , returns to
flying per euro spent are low and consumers would not fly if there were no purchasing opportunities at
destination: an increase in Π makes destination goods relatively more expensive and traveling becomes
less appealing. Vice-versa, when QI > TI , flying provides higher returns than home consumption itself:
while, again, destination goods become more expensive and are substituted with domestic consumption,
flying becomes relatively cheaper in real terms and its return per euro spent looms larger when other
prices increase.
Let’s now assume for a moment that (6) holds for both carriers and consider the choice of which
airline to patronize. Traveler k will fly with a Full Service carrier if
UFk ≥ ULk
⇐⇒
QF ≥
TF − TL
.
Π
(7)
Condition (7) is indeed a condition for the existence of the demand of Full Service flights and we assume
it holds throughout.20 Its interpretation is straightforward: as long as the price premium associated
18
See footnote 17.
It is easy to show that, for τ > 1 and P > 0, it holds Π ∈ (0, 1). See Lemma 2 for further details.
20
We will check later that, given the exogenous value of QF , equilibrium tariffs are consistent with condition (7).
19
8
to higher quality, TF − TL , is not too large, travelers prefer a more comfortable flight. However, the
condition is harder to be satisfied when the price index is small: in fact, when the price index is low, the
marginal utility of money spent at destination is high. Hence, the Full Service carrier has to increase
its quality or reduce its fare to attract travelers. Indeed, if prices are low enough, travelers may prefer
to save money on flight’s extra quality for goods’ consumption at destination. Thus, a quality premium
which is acceptable for high shop prices may become unacceptable when prices are low enough: flight
quality and goods’ purchases are substitutes.
Notice that assumption (7) does not depend on income: if it holds, whenever they can afford it,
travelers prefer to fly with a Full Service airline. Moreover, it implies that the income level satisfying the
flying condition (6) is lower for the Low Cost carrier than for the Full Service carrier. This gives rise to
a simple assignment of optimal flight choices based on income: low income consumers (with Ik <
TL
1−Π )
prefer to stay home; those with higher income want to travel and wish to fly with the Full Service airline,
but a part of them — say the middle class (with
TL
1−Π
< Ik < TF ) — cannot afford it, while the remaining
consumers — the upper class (with Ik > TF ) — fly with the high quality company. The market demand
for flights just described is illustrated in Figure 1.
Figure 1: Optimal flight choices and Income
No Fly
Ik = 0
Fly with L (DL )
Fly with F (DF )
TL
1−Π
TF
Ik = 1
While the Full Service carrier’s demand is fully determined by a mere cash constraint, the lower bound
of the Low Cost carrier’s demand is pinned down by a preference constraint and depends in a more
nuanced fashion on the goods’ consumption preferences through the price index Π. As it will be clear
later, this has deep implications as to the pricing policies and to the capability of the Low Cost carrier of
exploiting its platform nature. Finally, throughout the analysis we denote the demand for Full Service
and Low Cost flights respectively as DF ≡ 1 − TF and DL ≡ TF −
TL
1−Π .
Shops’ pricing. We now proceed to analyze the optimization problem of destination shops, which
engage in a standard monopolistic competition framework. First notice that demand for shop i, given
optimal consumer choices (4) and the income distribution, is
qi =
where
I˜ ≡
Z
P σ−τ
I˜
σ
pi 1 + P 1−τ
1
Z
(Ik − TF ) dIk +
TF
9
(8)
TF
TL
1−Π
(Ik − TL ) dIk
(9)
is travelers’ aggregate net income, i.e. the income not spent by travelers in flight tickets which is available
for purchasing goods. The overall demand for shop i, qi , decreases as pi increases while it increases when
˜ increases. The latter, in turn, clearly increases when the ticket of
the overall passengers’ net income, I,
the Low Cost airline, TL , decreases, as more passengers can afford to travel (extensive margin) and those
who already fly save some money (intensive margin); as to the effect of a change in the fare of the Full
Service airline, it depends on the uniform density of income in the following way: differentiating (9), a
decrease in TF increases shops’ demand if and only if 1 − TF > TF − TL ; put differently, the Full Service
tariff which maximizes net income is exactly midway between the maximum available income and the
Low Cost fare.
Shop i sets price pi under monopolistic competition to maximize profits
max {(pi − c) qi − sF (1 − TF ) − sL (TF − TL ) + zF + zL } ,
(10)
pi ≥0
where qi is given by 8 and sF and sL are the per–passenger transfers (possibly) agreed to with the Full
Service and the Low Cost airline respectively. The optimal price is the same for all shops and equals
p=
σ
c.
σ−1
(11)
Equation (11) has a simple interpretation: because shops have identical costs and traveler’s utility is
symmetric with respect to good varieties, the price is the same for all shops and is increasing in the
marginal cost. Moreover, p decreases and approaches the marginal cost as σ increases, i.e. Monopolistic
competition hits harder on shop’s margins as the elasticity of substitution between their products grows
larger, driving prices down to the competitive level.
Clearly, a shop’s profit is the same across shops and, given the optimal price and transfers sF ≥ 0
and sL ≥ 0, is
K I˜
TL
π=
− sF (1 − TF ) − sL TF −
+ zF + zL ,
ΠM
1−Π
(12)
where
K ≡ σ −1 P 1 + P τ −1
τ
1−τ
.
We summarize the subgame equilibrium play in the consumer goods’ market in the following lemma.
Lemma 1. Assume that: (i) condition (7) is satisfied; (ii) 0 < TL < TF < 1; and (iii) the contracts
(sL , zL ) and (sF , zF ) are accepted by the shops. Then:
• Consumers with income below
TL
1−Π
don’t fly; those with income between
TL
1−Π
and TF fly with the
Low Cost airline; and those with income above TF fly with the Full Service carrier.
• Destination shops charge the same price, p =
σ
σ−1 c
and each shop’s profit is (12).
• Non-traveling consumers spend all their income on the numéraire and their utility is UkN = Ik ;
10
passenger k flying with airline I purchases the quantity q0k =
the quantity qik =
q0k
σ
P τ M σ−1
Ik −TI
1+P 1−τ
of the numéraire good and
of all other goods i = 1, ..., M . The utility of travelers is (5).
Two sidedness. We have characterized the optimal decisions of travelers and shops, given carrier’s
strategies. The analysis has been standard up to this point. However, before studying airlines’ decisions,
we shall notice that our model has a distinct two-sided framework. Indeed, airlines in our model are
platforms connecting consumers on one side to shops on the other. A common definition of two-sided
markets — see Rochet and Tirole (2006) — is that, cœteris paribus, the net utility of players on one side
increases with the number of players on the other.21 In our setting, this amounts to say that, given TI
and QI such that some consumers travel, assuming that contracts sI and zI are accepted by the Mall,
and taking optimal purchasing and pricing decisions of travelers and shops, a traveler’s utility increases
with the number of shops at destination, M , and a shop’s profit increases with the number of travelers
arriving at destination.
To show that our model is indeed a two-sided market, notice that the number of consumers arriving
at destination is 1 − λ with λ ≡
TL
1−Π .
Thus, the market we are modeling can be defined a two-sided
market if a traveler’s utility (5) increases with M and a shop’s profit (12) gross of the payments to and
from the carrier(s) decreases with λ.
Substituting optimal price (11) into (5) yields the following utility for a traveler k who flies — i.e.
with income Ik ≥
TL
1−Π
— and profit for any shop
UIk = QI + (Ik − TI ) 1 + M
π =
τ −1
σ−1
σ−1
cσ
1
τ −1 ! τ −1
∀ I = L, F
K I˜λ
.
ΠM
where I˜λ is the aggregate net income (9) rewritten as a function of λ, that is
I˜λ ≡
Z
1
Z
TF
(Ik − TF ) dIk +
TF
(Ik − TL ) dIk .
λ
It is easy to see that a traveler’s utility increases with M , while a shop’s profit increases as the number
of travelers increases — i.e., λ decreases. We summarize the previous discussion in the next proposition.
Proposition 1. The flights market is characterized by network externalities typical of a two-sided market. In particular, for any profile of tariffs TI , qualities QI and transfers sI and zI , I = L, F , such
21
Another definition — see, again, Rochet and Tirole (2006) — is that the volume of transactions depends on how the
total price paid to the platform is shared between sides. Here it is clearly so: suppose that carrier I deals with the Mall
and charges tariff TI to travelers and SI per traveler to the Mall — i.e., a total platform price of TI + SI . It is clear from
(9) and Figure 1 that the volume of transactions I˜ as well as the overall flights’ demand 1 − TL /(1 − Π) are affected by
a change in the composition of the total platform price, either directly (under Low Cost – Mall contracting) or through
strategic interaction (under Full Service – Mall contracting).
11
that some consumers travel and shops agree to contracts, travelers gain from a larger number of shops
at destination, M , and shops profit from a larger number of travelers, 1 − λ.
We have shown that, in fact, airlines in our model are platforms of a two-sided market where the
sides to be connected are travelers/consumers and destination shops. Notice, however, that carriers are
not pure platforms whose value to a buyer (traveler) is only determined by the number of potential
sellers (shops) it provides a connection with: in fact (i) both carriers connect travelers to the same set
of shops; and (ii) buyers in our setup care about the platform quality, QI , which is independent from
the utility they derive from destination consumption. As we will show, this allows us to characterize
an equilibrium with vertical differentiation where the intrinsic value of the flight plays a distinct role,
neatly distinguishing our model from those of Argenziano (2008) and Gabszewicz and Wauthy (2014).
In the next sessions we will develop the analysis of airlines’ pricing strategies and characterize optimal
contracts between the airlines and the Mall.
22
Let’s thus denote the Mall’s profit as
K˜
TL
π S = I − SF (1 − TF ) − SL TF −
+ ZF + ZL ,
Π
1−Π
where the subscript
S
denotes the Mall. Note that
K
Π
is the profit per unit of net income if SI = ZI = 0
— i.e., when the Mall does not cooperate with any airline. Consistently, it takes values between zero
and one as proved by the following lemma.
Lemma 2. For all admissible values of the relevant parameters, i) 0 < K < Π < 1, and ii) Π < P .
Lemma 2 is useful because, introducing the parametric index K, it allows us to present most results
avoiding exponential functions, difficult to interpret, sign and visualize to the reader. Instead, most of
our results from now on will be expressed in terms of K and Π, which are conveniently ranked.
4
Market structures
We are interested in studying the viability of commercial agreements between airlines and destination
businesses. We will then characterize the optimal contracts proposed by different carriers to the Mall and
show how our model closely reproduces the special real world contracts typical of the flights industry.
Notice, however, that any contract can always be rejected by the Mall if it finds it unprofitable. Put
differently, the Mall has the possibility to implement standard competition between the carriers by
refusing to sign any cooperation agreement, a reference point carriers should consider when proposing
cooperation agreements with local businesses. Hence, in what follows we characterize first the standard
competition benchmark.
22
While shops’ optimization was clearly done at the shop-level, because shops are identical it is theoretically plain to
assume that they coordinate perfectly when dealing with airlines. Hence, we shift the focus of our analysis from the single
shop to the collective entity. See also footnote 15.
12
4.1
Standard competition benchmark
Denote by a superscript
Mall as
πC
i ,
C
the benchmark scenario, and the profits of the Low Cost, Full Service and
i ∈ {L, F, S}. It holds by assumption SIC = ZIC = 0, while optimal tariffs are chosen
simultaneously to maximize
πC
=
T
TF −
L
L
TL
1−Π
and
πC
F = TF (1 − TF ) ,
and
TFC = 21 .
yielding
TLC =
1−Π
4
Note that TLC , TFC is an equilibrium only if (7) is satisfied — i.e.,
QF ≥ QC
F,
with
QC
F ≡
1+Π
4Π .
(A1)
Finally, equilibrium profits in the competition benchmark are all positive and π C
L is decreasing in Π (see
Table 1 in the Appendix). This is because flying with the Low Cost airline increases travelers’ utility
just indirectly, by allowing them to purchase destination goods: as these become more expensive, the
carrier reduces its fare to avoid loosing too many travelers, and the final effect is a net loss to the airline.
We summarize the above in the following proposition.
Proposition 2. Assume SIC = ZIC = 0. Then, the flight industry equilibrium exists and is characterized
by tariffs TLC =
1−Π
4
and TFC =
1
2
and by a quality level QF ≥
1+Π
4Π .
Equilibrium demand for flights and
goods are those of Lemma 1.
We close this subsection by noting that any agreement between an airline and the Mall, which we
will analyze next, must be Pareto improving on the standard competition outcome for the cooperating
parties.
4.2
Low Cost – Mall agreement
Equipped with the benchmark case, we now characterize optimal tariffs TF and TL and transfers SL and
ZL when the Low Cost airline is the only carrier cooperating with the Mall — i.e. when SF = ZF = 0
by assumption. Denoting by a superscript
L
the case at hand, the Low Cost carrier sets TLL to jointly
L
maximize its own profits and those of the Mall, i.e. π L = π L
S + π L . He will then extract all the possible
surplus from the Mall through the two-part tariff SLL , ZLL where SLL is the per–passenger transfer while
ZLL is the fixed part. The airlines choose simultaneously TLL and TFL to maximize, respectively, the
following profits
πL =
K˜
ΠI +
TL TF −
TL
1−Π
13
and
πL
F = TF (1 − TF ) ,
where SLL and ZLL wash out of profits π L as they are mere transfers between the carrier and the Mall.
Optimal fares are
TLL =
(1−Π)2 (Π−K)
2(Π+(1−2Π)(Π−K))
and
TFL = 12 ,
where, by Lemma 2, TLL < TFL . Assuming for a moment that (7) is satisfied, let’s turn to the optimal
choice of SLL : the Low Cost carrier wants to implement the first best and will set SLL such that, given
L
L
TFL , the fist order condition when maximizing π L
L with respect to TL evaluated at TL is zero. Thus SL
solves
∂
(TL +SLL )
TL
TFL − 1−Π
∂TL
= 0,
which yields:
SLL =
K(1−Π)
2(Π+(1−2Π)(Π−K)) .
TL =TLL
Finally, TLL , TFL , SLL are equilibrium strategies only if (7) is satisfied, i.e.
QF ≥ QL
F,
with
QL
F ≡
1−Π(Π−K)
2(Π+(1−2Π)(Π−K)) .
(A2)
C
Notice that QL
F > QF : it is now more difficult to induce travelers to choose a full service flight. In
fact, due to the subsidy SLL payed by the Mall to the Low Cost carrier, the latter can reduce its tariff
(TLL < TLC ) thereby inducing a stronger preference for low cost flights, which has to be compensated
with higher flight quality from the full service airlines.
Before characterizing the optimal fixed transfer ZLL , let’s focus on profits, which are reported in Table
L
C
C
1. Given Lemma 2, it is not difficult to show that π L
L > π L and π F = π F for all values of the price
indexes — i.e., the agreement is beneficial to the Low Cost carrier while it leaves the Full Service carrier
indifferent vis-à-vis the competition benchmark. While the former result is intuitive insofar as the Low
Cost carrier enjoys one more instrument, SLL , than under standard competition, the latter depends on the
fact that, as discussed above, the Full Service carrier’s demand is only determined by a cash constraint,
which, in turn, is not affected by the agreement between the Low Cost airline and the Mall. Furthermore,
C
the zero cost assumption implies that the higher quality provided by the Full Service (QL
F > QF ) has
L
no impact on his profits. If we assumed a cost for quality then the natural result (π C
F > π F ) obtains.
C
As to the Mall’s profits, it can be shown that π L
S > π S if and only if Π > 5/6: in other words, given the
transfer required by the airline, the Mall’s profits are larger under cooperation when the prices charged
at destination are large enough.23
We can now complete the characterization of the optimal agreement between the Low Cost airline
L
and the Mall. The fixed part of the contract is clearly ZLL = π C
S − π S whenever positive and zero
otherwise — i.e., it amounts to a transfer from the airline to the Mall when the global price index falls
short of 5/6 and equals zero when prices are higher. The intuition is simple: when prices are high the
subsidy SLL is more than repaid by travelers’ purchases and the Mall can keep the additional surplus
because of the opt-out possibility. When, instead, prices are low, after paying the per-passenger subsidy,
the Mall is worse-off vis-à-vis the standard competition scenario and the airline has to compensate it
23
More precisely, Π > 5/6 if and only if P > 5 6τ −1 − 5τ −1
14
1
1−τ
1
, with P = M 1−σ
σ
c
σ−1
in equilibrium.
with a transfer ZLL > 0.
We have then established the following proposition.
Proposition 3. Assume SFL = ZFL = 0. Then, the flight industry equilibrium exists and is characterized
by tariffs TLL =
(1−Π)2 (Π−K)
2(Π+(1−2Π)(Π−K))
and TFL =
1
2,
and by a quality level QF ≥
1−Π(Π−K)
2(Π+(1−2Π)(Π−K)) .
The
equilibrium contract between the Low Cost carrier and the Mall prescribes a per-passenger transfer of
SLL =
K(1−Π)
2(Π+(1−2Π)(Π−K))
L
and, only if Π < 5/6, a fixed transfer ZLL = π C
S − π S Equilibrium profits are
given in Table 1. Equilibrium demand for flights and goods are those of Lemma 1.
4.3
Full Service – Mall agreement
We now turn to the case of a Full Service airline cooperating with the Mall. In this case, adopting a
similar notation, we assume SLF = ZLF = 0 and characterize the optimal contract SFF , ZFF proposed by
a Full Service carrier who maximizes joint profits π F = π FS + π FF . Profits at the stage of setting tariff are
π FL = TL TF −
TL
1−Π
πF =
and
K˜
ΠI +
TF (1 − TF ) ,
and the optimal fares are
Π−K
TLF = (1 − Π) 3(Π−K)+Π(1−K)
and
Π−K
TFF = 2 3(Π−K)+Π(1−K)
,
which, again by Lemma 2, are correctly ranked and properly define the demand function of Figure 1.
The optimal subsidy SFF is obtained as before and equals
SFF =
K(1−Π)
3(Π−K)+Π(1−K) .
Finally, there is demand for Full Service flights only if (7) is satisfied, i.e.
QF ≥ QFF ,
with
QFF ≡
Π−K
1+Π
Π 3(Π−K)+Π(1−K) .
(A3)
It is worth noticing that QFF < QC
F : it is easier for the Full Service carrier to attract travelers when it
cooperates with the Mall than under standard competition. In fact, because of the Mall’s subsidy SFF ,
the Full Service airline can reduce its tariff (TFF < TFC ) and this reduction is proportionally larger than
the strategic reduction in flight fare of the Low Cost airline (TLF < TLC ), inducing travelers to prefer a
full service airline at lower quality levels.
F
C
Profits are reported in Table 1. Using Lemma 2 it can be easily shown that π FF > π C
F , πL < πL
and π FS < π C
S for all values of the price indexes, i.e. cooperation with the Mall benefits the Full Service
carrier to the expenses of both the Low Cost and the Mall. The first inequality is, again, quite intuitive.
The second has to do with the fact that now the reduction of the Full Service carrier’s fare erodes the
demand of the Low Cost airline from above, causing a loss in equilibrium. Finally, the Mall’s profits are
negatively affected no matter the level of Π.
15
We can now complete the characterization of the optimal agreement between the Full Service airline
F
and the Mall. The fixed part of the contract is clearly Z F = π C
S − π S and amounts to a transfer from
the airline to the Mall. We have then established the next proposition.
Proposition 4. Assume SLF = ZLF = 0. Then, the flight industry equilibrium exists and is charΠ−K
Π−K
and TFF = 2 3(Π−K)+Π(1−K)
, and by a quality level
acterized by tariffs TLF = (1 − Π) 3(Π−K)+Π(1−K)
QF ≥
Π−K
1+Π
Π 3(Π−K)+Π(1−K) .
The equilibrium contract between the Full Service carrier and the Mall pre-
scribes a per-passenger transfer of SFF =
K(1−Π)
3(Π−K)+Π(1−K)
F
and a fixed transfer ZFF = π C
S − π S , where
equilibrium profits are given in Table 1. Equilibrium demand for flights and goods are those of Lemma 1.
4.4
Comparing market structures
We now compare the market structures just characterized along several dimensions. The next corollary
summarizes results discussed in the Sections 4.2 and 4.3.
Corollary 1. For all admissible values of c, ρ, ϕ and M the next inequalities hold:
TLC > TLF > TLL ,
TFC = TFL > TFF ,
C
F
QL
F > QF > QF ,
C
F
πL
L > πL > πL ,
L
π FF > π C
F = πF .
Most intuitions behind Corollary 1 have been discussed above. We remark here that the tariffs of both
carriers are highest in the benchmark scenario, when no agreements are signed. In fact, when a carrier
strikes a deal with the Mall, the latter subsidizes the former so as to allow the carrier to charge lower
tariffs and embark a larger number of potential customers. Absent a deal with the Mall, the carrier
does not internalize this ‘platform externality’ and charges higher tariffs which discourage consumers
from traveling. This has neat implications for the flights’ demand — i.e., the number of travelers — as
detailed in the next corollary.
Corollary 2. For all admissible values of c, ρ, ϕ and M the next inequalities hold:
i
DFi > DL
i ∈ {C, L, F } ;
DFF > DFC = DFL ;
L
C
F
DL
> DL
> DL
;
L
F
C
DL
+ DFL > DL
+ DFF > DL
+ DFC .
The first inequality confirms a standard vertical differentiation result: the quality leader enjoys a larger
market share. And this is true regardless of the market structure: the Full Service carrier has higher
demand than the Low Cost carrier. The second and third inequalities confirm the basic intuition that
cooperation with the Mall allows a carrier to reduce fares and thereby increase his market share of flights.
The last inequality states that overall flights’ demand increases when a carrier deals with the Mall and
is maximized when it is the Low Cost carrier to do so.
16
5
Endogenous market structure
In Section 3 we have characterized the equilibrium of the flights industry taking the market structure
as given. We now relax this assumption allowing the market structure to be determined endogenously.
In particular, we assume that contracts between an airline and the Mall are signed under exclusivity —
i.e. SL · SF = ZL · ZF = 0. This assumption is natural on two grounds: first, under non exclusivity
carriers would clearly form a cartel to maximize the grand coalition profits, triggering the intervention of
antitrust authorities; second, the type of contracts Ryanair has on its bases are never observed between
an airport and more than one carrier. Hence, the Mall chooses which contract to accept, if any, between
the two offered by the airlines.
We wish to understand which market structure emerges in equilibrium among those characterized
and how this depends on the relevant parameters of the model. Propositions 3 and 4 have proved that
cooperation with either airline always increases the joint profits of the cooperating parties vis-à-vis the
standard competition scenario. Hence, in the equilibrium market structure the Mall cooperates with a
carrier. Clearly, the Mall chooses the carrier which guarantees the largest increase in total profits to the
cooperating parties: in fact, cooperation with the other airline generates a smaller increase in profits and
is vulnerable to a proposal from the rival airline to the Mall which, trough appropriate side payments,
would induce the Mall to switch partner.
C
L
Denote by ∆L
L = π L − π L the profit change for airline L between the Low Cost–Mall cooperation
F
F
scenario and the competition benchmark, and define similarly ∆L
S , ∆F and ∆S , where the subscript
identifies the player and the superscript the market structure — closed form expressions are in Table 2.
L
F
F
Then, writing ξ (Π, K) ≡ ∆L
and relegating to Table 3 in the Appendix its
L + ∆ S − ∆F + ∆ S
cumbersome expression, it is clear that cooperation with the Low Cost (resp. Full Service) is the
equilibrium outcome if, and only if, ξ (Π, K) > 0 (resp. <). The following proposition gives conditions
under which ξ (Π, K) is greater or smaller than zero and characterizes the equilibrium market structure.
Proposition 5. Assume that cooperation contracts between the Mall and airlines are exclusive. Then,
the optimal market structure of the airline industry features cooperation between the Mall and the Low
Cost carrier and the optimal contract is characterized in Proposition 3 if, and only if, σ, τ , c and M are
such that
K < f (Π)
(13)
where f (Π) — illustrated in Figure 2 and reported in Table 3 — is increasing in Π and is such that
f (Π) < Π for all Π lower than 1, with f (Π) = 0 for Π ≈ 0.24512.24 If condition (13) holds with the
opposite sign, then the optimal market structure is cooperation between the Mall and the Full Service
carrier and the optimal contract is characterized in Proposition 4.
24
A closed form expression for Π such that f (Π) = 0 cannot be obtained. Indeed, from now on, some of the proofs will
be graphical and/or will use computational software.
17
1
K
f HPL
0
P
0.24512
1
Figure 2: Market structure and parameter regions.
Full Service – Mall;
Low Cost – Mall.
Hence, the equilibrium market structure is cooperation with the Full Service carrier if σ, τ , c and M
◦
are such that (K, Π) falls in the area between the 45 line and f (Π), while cooperation with the Low
Cost airline is the equilibrium outcome in the opposite case.
It is apparent that the area corresponding to the market structure in which the Low Cost carrier
cooperates with the Mall — i.e., what happens in most real world contracts — is much larger than the
area corresponding to cooperation between the Mall and the Full Service carrier — which, in fact, is
rarely observed. However, because K and Π are complex functions of the parameters of the model, it is
difficult to properly interpret Figure 2.
To provide more intuition, we perform some comparative statics on the parameters σ, τ , c and M
and study their implications at the locus of points implicitly defined by K = f (Π). To understand the
contribution of the underlying parameters to the equilibrium outcome we run the following exercise: we
let parameters vary one by one and show that, for each of them, K and Π move in opposite directions.
This allows us to draw neat conclusions on which market structure emerges when (K, Π) is located on
the edge K = f (Π) and a parameter changes. To see this, suppose that we increase a parameter and
this causes K to decrease and Π to increase (resp. K to increase and Π to decrease): then we can say
that the set of the remaining parameters under which cooperation between the Mall and the Low Cost
airline is the equilibrium market structure is larger (resp. smaller).
The next proposition illustrates the relevant comparative statics of the model.
1
Proposition 6. Assume that M ≥ 3 and P < e− e . Then K (resp. Π) is decreasing (resp. increasing)
in c, ρ and ϕ and increasing (resp. decreasing) in M .
Proposition 6 implies that, loosely speaking, cooperation between the Mall and, say, the Low Cost
carrier — i.e., K < f (Π) — is more likely the higher the production cost c of destination goods and
18
the higher the degree of substitution between destination goods, ρ, and between home and destination
consumption, ϕ. Cooperation with the Full Service is instead more likely when the number of shops at
destination is larger.
A higher marginal cost of production of destination shops pushes up the retail price of destination
goods, p = ρc , increasing both the destination (P ) and the global (Π) price indexes. Through the carriers’
tariffs, this affects consumers’ incentives to travel.25,26 In particular, the number of travelers diminishes
in all market structures when prices increase. However, when the Low Cost carrier cooperates with the
Mall, the demand of high quality flights is not affected while that of low cost flights diminishes.27 Vice
versa, the demand of high quality flights diminishes while that of low cost flights increases when the Full
Service carrier cooperates with the Mall. In fact, when the Mall cooperates with the Low Cost, the Full
Service does not internalize the downstream price increase. The opposite is true if the Full Service deals
with the Mall: a higher price of destination goods invites the carrier to charge a higher tariff. Because
consumers prefer high quality flights if they can afford them (see condition (7)), this asymmetry implies
that, under Low Cost–Mall cooperation the negative impact of a price increase on the travelers’ spending
power is attenuated because the Full Service carrier does not revise upward its tariff. In other words, by
cooperating with the Low Cost carrier, the Mall minimizes the impact of higher prices on the spending
power of travelers because the Full Service does not respond by increasing tariffs under Low Cost – Mall
contracting. Hence, the higher the marginal cost c, the more convenient it is for the Mall to deal with
the Low Cost carrier.
Changes in M , ρ and ϕ do not affect retail prices at destination. They rather impact a consumer’s
utility through the price indexes P and Π. These, by the properties of CES utility functions, denote the
cost of a unit of utility.28 An increase in the number of shops, M , makes dealing with the Full Service
carrier more convenient. The intuition is as follows: because consumers love variety (ρ < 1 and ϕ < 1),
the larger the number of different products, M , the larger the utility a consumer attains with a given
budget. Or, which is equivalent, the lower the cost of a unit of utility. Hence, a higher M implies a
lower destination price index P and, a fortiori, a lower global price index Π. This, in turn, increases
overall flights’ demand and triggers: i) a tariff reduction and demand increase of high quality flights and
a decrease of low cost flights’ demand under Full Service–Mall cooperation; ii) an increase in low cost
flights’ demand under Low Cost–Mall contracting. It turns out that the gains under Full Service–Mall
contracting at the intensive margin — more money to spend because of a lower tariff — are larger than
25
The sign of derivatives w.r.t. P is the same of those w.r.t. Π as both indexes co-move.
Note that, for j = L, F , it holds ∂TFj /∂Π > 0, while ∂TLj /∂Π can be either positive or negative depending on parameters.
In particular, TLj increases with Π if K/Π is sufficiently high — i.e., when the shops’ profitability is sufficiently large.
Moreover, the per–passenger transfer from the Mall to the cooperating carrier increases with the price index in both market
structures — i.e., ∂Sjj /∂Π > 0
27
In fact, DFL = 21 , while Dij is a function of Π and K in all other cases (i, j) 6= (F, L).
28
More precisely, P is the monetary cost of a unit of utility derived from the equilibrium consumption of a bundle of
destination goods, while Π is the monetary cost of a unit of utility derived from the equilibrium consumption of a composite
bundle of the home (numéraire) and destination goods. This property derives from the homogeneity of degree 1 of CES
functions.
26
19
the loss at the extensive margin — more travelers fly with the Full Service than with the Low Cost
carrier at a tariff higher than that of cheap flights — and that the net gains are larger than the gains at
the extensive margin
29
under Low Cost–Mall cooperation. Hence, reverting the argument made above,
an increase in the number of varieties of destination goods makes dealing with the Full Service carrier
relatively more convenient: consumers derive a larger utility from the same money spent at destination
and optimally substitute some shopping with higher quality flights. Put differently, the need to spare
money on flights is less acute when shopping at destination avails consumers a higher utility.
Finally, an increase in ρ and/or in ϕ makes goods less differentiated in the consumer’s eyes.30 Hence,
consumers derive less utility from a given products’ bundle or, equivalently, the price of utility increases
— i.e., P and Π increase. A higher ρ has the additional effect of reducing the spot price of destination
goods, p = ρc . This is a natural consequence of monopolistic competition between shops at destination:
the more their products substitute one another, the tougher is competition, the lower are prices. This
pushes P down. However, the net effect of a higher ρ on the price index P is positive. The increase
in P and Π following a drop in the love for variety makes consumers eager to spend more money on
goods’ purchases saving on flights quality to balance their overall consumption choices. While, again,
overall flights’ demand drops, cooperation with the Low Cost carrier allows the Mall to contain the loss
on the intensive margin that would be caused by the Full Service carrier raising his tariff under Full
Service–Mall cooperation. Hence, cooperation with the Low Cost carrier is more efficient when products
become more substitutes.
6
Conclusion
We have developed a simple model of the air travel industry which rationalizes the extremely low tariffs
practiced by the (ultra) low cost carrier Ryanair. Our model does so by recognizing that Ryanair exploits
the two-sided nature of air traveling: bringing potential buyers closer to potential sellers. Pricing both
sides of the market rather than focusing just on cost containment and quality competition — as standard
low cost carriers in regular vertically differentiated markets — Ryanair has been able to further reduce
the cost of flying to potential travelers inducing higher demand and allowing lower income people to
travel. These findings of the model correspond to facts that have been observed in practice. Further, we
show that it is more likely that low cost carriers adopt platform pricing rather than full service carriers
whenever deriving utility from goods’ consumption becomes more costly so that saving on flight’s quality
become more important — e.g., because the production cost of destination goods increases or due to
changes in preferences. To the contrary, we show that it is relatively more likely that full service carriers
use platform pricing whenever travelers derive more utility spending a given budget on destination goods
— e.g., because their variety increases making destination consumption more appealing.
A decrease in Π may cause a gain as well as a loss at the intensive margin. This is because the sign of ∂TLj /∂Π, j = L, F
depends on the specific values of the parameters.
30
The love for variety diminishes and the elasticities of substitution σ and τ increase.
29
20
References
Argenziano, R. (2008): “Differentiated Networks: Equilibrium and Efficiency,” The RAND Journal
of Economics, 39(3), 747–769.
Armstrong, M. (2006): “Competition in two-sided markets,” The RAND Journal of Economics, 37(3),
668–691.
Dixit, A. K., and J. E. Stiglitz (1977): “Monopolistic Competition and Optimum Product Diversity,” American Economic Review, 67(3), 297–308.
Gabszewicz, J. J., and X. Y. Wauthy (2014): “Vertical product differentiation and two-sided markets,” Economic Letters, 123, 58–61.
Ribeiro, V. M., J. Correia-da-Silva, and J. Resende (2014): “Nesting vertical and horizontal
differentiation in two-sided markets,” CEF working paper.
Rochet, J.-C., and J. Tirole (2006): “Two-sided markets: a progress report,” The RAND Journal
of Economics, 37(3), 645–667.
Tirole, J. (1988): The Theory of Industrial Organization, vol. 1 of MIT Press Books. The MIT Press.
21
Appendix
The missing proofs will be soon available at www.giovanniursino.eu/papers.
Marshallian demand:
∂
∂qik
θQIk +
X
M
qρ
i=1 ik
1
ρ
!
XM
pi qik
= 0
+ λ Ik − TI −
i=1
1
ρ
X
M
i=1
ρ
qik
1 −1
ρ
ρ−1
ρqik
− λpi = 0
X
M
qρ
i=1 ik
1−ρ
ρ
P
ρ−1
qik
= λpi
ρ
M
i=1 qik
1
qik
ρ
1
1
= λ 1−ρ pi1−ρ
P
ρ
M
i=1 qik
qik =
1
1
1
ρ
(Ax.1)
λ 1−ρ pi1−ρ
from which, calling y ≡
P
ρ
M
i=1 qik
1
ρ
, we can plug (Ax.1) into the budget constraint (2) and obtain
ρ
λ
1
1−ρ
PM
=
ρ−1
i=1 pi
y
Ik − TI
and, plugging this back into (Ax.1),
1
Ik − TI 1−ρ
qik = P
p
ρ
i
M
ρ−1
i=1 pi
which, substituting σ =
1
1−ρ ,
becomes
Ik − TI
qik = PM 1−σ p−σ
i .
i=1 pi
Q.E.D.
Proof of Lemma 2: Clearly K > 0, Π > 0 and P > 0.
Proof that K < Π. Suppose not. Then, substituting for K and Π it must be:

c (1 − ρ) 
ρM
1−ρ
ρ
1−
1+
c ρ−1
M ρ
ρ
ϕ
ϕ−1
!−1  ϕ1
 >
22
1+
c ρ−1
M ρ
ρ
ϕ
ϕ−1
! ϕ−1
ϕ
.
(Ax.2)
ϕ ρ−1 ϕ−1
c
ρ
both sides of (Ax.2). Simplify the
Take first the power of ϕ and then multiply by 1 + ρ M
left hand side and get
(1 − ρ)
Now thake the power of
1
ϕ
ϕ
c ρ−1
M ρ
ρ
ϕ2
ϕ−1
>
1+
c ρ−1
M ρ
ρ
ϕ
ϕ−1
!ϕ
.
(Ax.3)
on both sides of (Ax.3) and simplify to get the following condition
−ρ
c ρ−1
M ρ
ρ
ϕ
ϕ−1
> 1,
which is clearly impossible given the admissible parameter values. Hence, it must be K < Π.
Proof that Π < 1. Suppose not, then it must be Π−1 < 1. Substituting for Π, this implies
1+
Thake the (positive) power of
ϕ
1−ϕ
c ρ−1
M ρ
ρ
ϕ
ϕ−1
! 1−ϕ
ϕ
< 1.
(Ax.4)
on both sides of (Ax.4) and simplify to get
c ρ−1
M ρ
ρ
ϕ
ϕ−1
< 0,
which is clearly impossible. Hence, it must be Π < 1.
Proof that Π < P . Suppose not, then, substituting for P and Π, it must be
c ρ−1
M ρ <
ρ
Take the power of
ϕ
1−ϕ
1+
c ρ−1
M ρ
ρ
ϕ
ϕ−1
! ϕ−1
ϕ
.
(Ax.5)
< 1.
(Ax.6)
on both sides of (Ax.5) and rearrange to get
c ρ−1
M ρ
ρ
ϕ
1−φ
1+
c ρ−1
M ρ
ρ
Expand and simplify (Ax.6) to get
c ρ−1
M ρ
ρ
ϕ
1−φ
< 0,
which is never satisfied. Hence, it must be Π < P . Q.E.D.
23
ϕ
ϕ−1
!
Table 1: Equilibrium profits under different scenarios.
Comp.
Low Cost/Mall coop.
Full Service/Mall coop.
2
πL
1−Π
16
(1−Π)Π2 (1−Π+K)
4(Π+(1−2Π)(Π−K))2
(1−Π)(Π−K)2
(3(Π−K)+Π(1−K))2
πF
1
4
1
4
(K−2Π+KΠ)2
(3(Π−K)+Π(1−K))2
πS
K 2Π+5
Π 32
K
K(2Π−1)((Π−K)(1−Π(2+Π))+Π(3−Π2 ))+Π2 (2Π+3)(1−Π)2
8Π(Π+(1−2Π)(Π−K))2
K
Π2 (6Π+1)−(Π2 +4Π+2)(2Π−K)K
2Π(3(Π−K)+Π(1−K))2
Table 2: Profit differentials with respect to the competition benchmark.
i
∆L
i
L
(1−Π)(4Π−K)+3KΠ
K (1 − Π) 16(Π+(1−2Π)(Π−K))
2 > 0
F
0
S
∆Fi
−K (1 − Π)2
<0
5(Π−K)+3Π(1−K)
K (1 − Π) 4(3(Π−K)+Π(1−K))
2 > 0
2 (1−Π)2
2
7(Π−K)+Π(1−K)
16(3(Π−K)+Π(1−K))2
+4Π(1−2Π)(1−Π)K−8Π
K −(1−2Π)K32Π(Π+(1−2Π)(Π−K))
2
K (1 − Π) 2K
2 Π2 +56KΠ−64Π2 +16KΠ2 +3K 2 Π−13K 2
2
32Π(3(Π−K)+Π(1−K))
<0
Table 3: Joint profit differentials with respect to the competition benchmark and their difference.
L
∆L
L + ∆S =
K2
32Π(Π+(1−2Π)(Π−K))
>0
∆FF + ∆FS = K 2 (1 − Π) 8(Π−K)+(1−Π)(5(Π−K)+Π(3−2K))
>0
32Π(3(Π−K)+Π(1−K))2
ξ (Π, K) ≡ K
K (2K 2 (Π(Π3 −8Π+12)−1)+KΠ(9−Π(Π(Π(2Π+7)−45)+61))+8Π2 (Π((Π−4)Π+5)−1))
16Π(Π+(1−2Π)(Π−K))(3(Π−K)+Π(1−K))2
√
f (Π) ≡
3
2
3
4
Π 61Π−45Π +7Π +2Π −9− (1−Π) (41Π−5Π2 −41Π3 +24Π4 −4Π5 +17)
4
12Π−8Π2 +Π4 −1
24